Upload FIM-ODE fhn training config, logging, etc...
Browse files- fhn/gridsearch_01-29-1048/finetuning_report.pdf +0 -0
- fhn/gridsearch_01-29-1048/logging/tensorboard/events.out.tfevents.1769683690.ml2ran03.lamarr.tu-dortmund.de.926253.0 +3 -0
- fhn/gridsearch_01-29-1048/logging/tensorboard/events.out.tfevents.1769683691.ml2ran03.lamarr.tu-dortmund.de.926253.1 +3 -0
- fhn/gridsearch_01-29-1048/logging/train.log +0 -0
- fhn/gridsearch_01-29-1048/model_architecture.txt +0 -0
- fhn/gridsearch_01-29-1048/task_loss.log +97 -0
- fhn/gridsearch_01-29-1048/train_parameters.yaml +110 -0
fhn/gridsearch_01-29-1048/finetuning_report.pdf
ADDED
|
Binary file (10.7 kB). View file
|
|
|
fhn/gridsearch_01-29-1048/logging/tensorboard/events.out.tfevents.1769683690.ml2ran03.lamarr.tu-dortmund.de.926253.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d767da5a92fa8a507025bc9e2d0bd3ba611097edf6c51891494f0c323baa6ac
|
| 3 |
+
size 88
|
fhn/gridsearch_01-29-1048/logging/tensorboard/events.out.tfevents.1769683691.ml2ran03.lamarr.tu-dortmund.de.926253.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4a1a76c8b5f81f07695abcb09236f94edffac1843feaf016bead6b7d8f970a40
|
| 3 |
+
size 271365
|
fhn/gridsearch_01-29-1048/logging/train.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
fhn/gridsearch_01-29-1048/model_architecture.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
fhn/gridsearch_01-29-1048/task_loss.log
ADDED
|
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
201: 0.271
|
| 2 |
+
203: 0.071
|
| 3 |
+
205: 0.041
|
| 4 |
+
207: 0.109
|
| 5 |
+
209: 0.126
|
| 6 |
+
211: 0.081
|
| 7 |
+
213: 0.097
|
| 8 |
+
215: 0.040
|
| 9 |
+
217: 0.042
|
| 10 |
+
219: 0.104
|
| 11 |
+
221: 0.053
|
| 12 |
+
223: 0.055
|
| 13 |
+
225: 0.135
|
| 14 |
+
227: 0.148
|
| 15 |
+
229: 0.036
|
| 16 |
+
231: 0.032
|
| 17 |
+
233: 0.052
|
| 18 |
+
235: 0.064
|
| 19 |
+
237: 0.123
|
| 20 |
+
239: 0.057
|
| 21 |
+
241: 0.047
|
| 22 |
+
243: 0.066
|
| 23 |
+
245: 0.054
|
| 24 |
+
247: 0.082
|
| 25 |
+
249: 0.038
|
| 26 |
+
251: 0.049
|
| 27 |
+
253: 0.038
|
| 28 |
+
255: 0.047
|
| 29 |
+
257: 0.036
|
| 30 |
+
259: 0.045
|
| 31 |
+
261: 0.034
|
| 32 |
+
263: 0.041
|
| 33 |
+
265: 0.035
|
| 34 |
+
267: 0.052
|
| 35 |
+
269: 0.041
|
| 36 |
+
271: 0.061
|
| 37 |
+
273: 0.051
|
| 38 |
+
275: 0.062
|
| 39 |
+
277: 0.045
|
| 40 |
+
279: 0.064
|
| 41 |
+
281: 0.053
|
| 42 |
+
283: 0.074
|
| 43 |
+
285: 0.048
|
| 44 |
+
287: 0.090
|
| 45 |
+
289: 0.039
|
| 46 |
+
291: 0.079
|
| 47 |
+
293: 0.064
|
| 48 |
+
295: 0.040
|
| 49 |
+
297: 0.062
|
| 50 |
+
299: 0.048
|
| 51 |
+
301: 0.042
|
| 52 |
+
303: 0.035
|
| 53 |
+
305: 0.088
|
| 54 |
+
307: 0.045
|
| 55 |
+
309: 0.041
|
| 56 |
+
311: 0.076
|
| 57 |
+
313: 0.032
|
| 58 |
+
315: 0.043
|
| 59 |
+
317: 0.108
|
| 60 |
+
319: 0.037
|
| 61 |
+
321: 0.040
|
| 62 |
+
323: 0.087
|
| 63 |
+
325: 0.031
|
| 64 |
+
327: 0.049
|
| 65 |
+
329: 0.082
|
| 66 |
+
331: 0.048
|
| 67 |
+
333: 0.042
|
| 68 |
+
335: 0.037
|
| 69 |
+
337: 0.043
|
| 70 |
+
339: 0.046
|
| 71 |
+
341: 0.030
|
| 72 |
+
343: 0.042
|
| 73 |
+
345: 0.045
|
| 74 |
+
347: 0.075
|
| 75 |
+
349: 0.053
|
| 76 |
+
351: 0.036
|
| 77 |
+
353: 0.041
|
| 78 |
+
355: 0.039
|
| 79 |
+
357: 0.053
|
| 80 |
+
359: 0.043
|
| 81 |
+
361: 0.036
|
| 82 |
+
363: 0.043
|
| 83 |
+
365: 0.036
|
| 84 |
+
367: 0.038
|
| 85 |
+
369: 0.056
|
| 86 |
+
371: 0.032
|
| 87 |
+
373: 0.029
|
| 88 |
+
375: 0.046
|
| 89 |
+
377: 0.039
|
| 90 |
+
379: 0.055
|
| 91 |
+
381: 0.038
|
| 92 |
+
383: 0.050
|
| 93 |
+
385: 0.044
|
| 94 |
+
387: 0.040
|
| 95 |
+
389: 0.030
|
| 96 |
+
391: 0.032
|
| 97 |
+
393: 0.037
|
fhn/gridsearch_01-29-1048/train_parameters.yaml
ADDED
|
@@ -0,0 +1,110 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
dataset:
|
| 2 |
+
batch_size:
|
| 3 |
+
test: 1
|
| 4 |
+
train: 1
|
| 5 |
+
validation: 1
|
| 6 |
+
data_dirs:
|
| 7 |
+
test: !!python/tuple []
|
| 8 |
+
train: !!python/tuple
|
| 9 |
+
- experiments/odeon/fhn/data_gpode
|
| 10 |
+
validation: !!python/tuple
|
| 11 |
+
- experiments/odeon/fhn/data_gpode
|
| 12 |
+
dataset_name:
|
| 13 |
+
test: HeterogeneousFIMSDEDataset
|
| 14 |
+
train: StreamingFIMSDEDataset
|
| 15 |
+
validation: StreamingFIMSDEDataset
|
| 16 |
+
files_to_load:
|
| 17 |
+
locations: locations.h5
|
| 18 |
+
obs_mask: obs_mask.h5
|
| 19 |
+
obs_times: obs_times.h5
|
| 20 |
+
obs_values: obs_values.h5
|
| 21 |
+
max_dim: 3
|
| 22 |
+
name: FIMSDEDataloaderIterableDataset
|
| 23 |
+
num_locations:
|
| 24 |
+
test: null
|
| 25 |
+
train: null
|
| 26 |
+
validation: null
|
| 27 |
+
num_observations:
|
| 28 |
+
test: null
|
| 29 |
+
train: null
|
| 30 |
+
validation: null
|
| 31 |
+
num_workers:
|
| 32 |
+
test: 0
|
| 33 |
+
train: 1
|
| 34 |
+
validation: 1
|
| 35 |
+
shard:
|
| 36 |
+
test: false
|
| 37 |
+
train: false
|
| 38 |
+
validation: false
|
| 39 |
+
shuffle_elements: true
|
| 40 |
+
shuffle_locations:
|
| 41 |
+
test: false
|
| 42 |
+
train: false
|
| 43 |
+
validation: false
|
| 44 |
+
shuffle_paths: true
|
| 45 |
+
distributed:
|
| 46 |
+
activation_chekpoint: false
|
| 47 |
+
checkpoint_type: full_state
|
| 48 |
+
enabled: false
|
| 49 |
+
min_num_params: 1e5
|
| 50 |
+
sharding_strategy: NO_SHARD
|
| 51 |
+
wrap_policy: SIZE_BAZED
|
| 52 |
+
experiment:
|
| 53 |
+
device_map: auto
|
| 54 |
+
name: gridsearch
|
| 55 |
+
name_add_date: true
|
| 56 |
+
seed: 10
|
| 57 |
+
model:
|
| 58 |
+
model_config:
|
| 59 |
+
attention_map: softmax
|
| 60 |
+
attention_method: linear
|
| 61 |
+
dim_embed: 256
|
| 62 |
+
dim_feedforward: 1024
|
| 63 |
+
dim_ffn_u_model: 1024
|
| 64 |
+
dim_hidden_u_model: 256
|
| 65 |
+
dim_max_trajectory: 3
|
| 66 |
+
dropout: 0.0
|
| 67 |
+
num_context_encoder_layers: 2
|
| 68 |
+
num_heads: 8
|
| 69 |
+
num_res_layer_u_model: 6
|
| 70 |
+
num_res_layers_functional_decoder: 8
|
| 71 |
+
use_bias_for_projection: true
|
| 72 |
+
use_bias_in_attention: true
|
| 73 |
+
use_query_residual_in_attention: true
|
| 74 |
+
model_type: TrainingWrapper
|
| 75 |
+
train_config:
|
| 76 |
+
corruption_model_type: null
|
| 77 |
+
h_max: null
|
| 78 |
+
ic_noise_scale: 0.0
|
| 79 |
+
integrator_for_trajectory_training: rk4
|
| 80 |
+
intermediate_steps_per_step: 5
|
| 81 |
+
loss_filter_nans: true
|
| 82 |
+
loss_type: l1
|
| 83 |
+
num_ic: 1
|
| 84 |
+
only_final_points_for_loss: false
|
| 85 |
+
step_noise_scale: 0.0
|
| 86 |
+
train_type: trajectory_reconstruction
|
| 87 |
+
train_with_normalized_head: true
|
| 88 |
+
traj_loss_steps: 19
|
| 89 |
+
use_h_max: false
|
| 90 |
+
optimizers: !!python/tuple
|
| 91 |
+
- optimizer_d:
|
| 92 |
+
gradient_norm_clipping: 1.0
|
| 93 |
+
lr: 1.0e-05
|
| 94 |
+
name: torch.optim.AdamW
|
| 95 |
+
weight_decay: 0.001
|
| 96 |
+
trainer:
|
| 97 |
+
best_metric: loss
|
| 98 |
+
debug_iterations: null
|
| 99 |
+
detect_anomaly: false
|
| 100 |
+
epochs: 401
|
| 101 |
+
experiment_dir: results/fhn
|
| 102 |
+
gradient_accumulation_steps: 1
|
| 103 |
+
logging_format: RANK_%(rank)s - %(asctime)s - %(name)s - %(levelname)s - %(message)s
|
| 104 |
+
name: Trainer
|
| 105 |
+
precision: bf16mixed
|
| 106 |
+
save_every: 1
|
| 107 |
+
schedulers: !!python/tuple
|
| 108 |
+
- beta: 1.0
|
| 109 |
+
label: drift_loss_scale
|
| 110 |
+
name: fim.utils.param_scheduler.ConstantScheduler
|